An algorithm for unsupervised linear discriminant analysis was presented. Optimal unsupervised discriminant vectors are obtained through maximizing covariance of all samples and minimizing covariance of local k-neares...
详细信息
An algorithm for unsupervised linear discriminant analysis was presented. Optimal unsupervised discriminant vectors are obtained through maximizing covariance of all samples and minimizing covariance of local k-nearest neighbor samples. The experimental results show our algorithm is effective.
Efficient reconfigurable VLSI architecture for 1-D 5/3 and 9/7 wavelet transforms adopted in JPEG2000 proposal, based on lifting scheme is proposed. The embedded decimation technique based on fold and time multiplexin...
详细信息
Efficient reconfigurable VLSI architecture for 1-D 5/3 and 9/7 wavelet transforms adopted in JPEG2000 proposal, based on lifting scheme is proposed. The embedded decimation technique based on fold and time multiplexing, as well as embedded boundary data extension technique, is adopted to optimize the design of the architecture. These reduce significantly the required numbers of the multipliers, adders and registers, as well as the amount of accessing external memory, and lead to decrease efficiently the hardware cost and power consumption of the design. The architecture is designed to generate an output per clock cycle, and the detailed component and the approximation of the input signal are available alternately. Experimental simulation and comparison results are presented, which demonstrate that the proposed architecture has lower hardware complexity, thus it is adapted for embedded applications. The presented architecture is simple, regular and scalable, and well suited for VLSI implementation.
An algorithm for unsupervised linear discriminant analysis was presented. Optimal unsupervised discriminant vectors are obtained through maximizing covariance of all samples and minimizing covariance of local k-neares...
详细信息
An algorithm for unsupervised linear discriminant analysis was presented. Optimal unsupervised discriminant vectors are obtained through maximizing covariance of all samples and minimizing covariance of local k-nearest neighbor samples. The experimental results show our algorithm is effective.
This paper presented a new algorithm for face detection in complex environment from images. The algorithm works by first doing lighting compensation on input image, then it segments and combines skin regions obtained ...
详细信息
This paper presented a new algorithm for face detection in complex environment from images. The algorithm works by first doing lighting compensation on input image, then it segments and combines skin regions obtained through applying skin model, and extracts face candidates with the aid of heuristic information. Finally, it evaluates the existence of facial features, such as face boundary, eyes and mouth, in those face candidates. The detection rate has achieved 89.7% when the algorithm is employed to detect 1010 color face images containing face rotations with complex environment and lighting variances.
The infomax algorithm was applied to perform the feature extraction for the ship-radiated noise. It is proved that the ICA transform can improve the sup-Gaussian property of ship-radiated noise. Based on the property ...
详细信息
The infomax algorithm was applied to perform the feature extraction for the ship-radiated noise. It is proved that the ICA transform can improve the sup-Gaussian property of ship-radiated noise. Based on the property of sparse coding, an efficient de-noising result can be obtained by the threshold method. The de-nosing experiments of ship-radiated noise with sea noise show that the proposed method is valid and more efficient than other conventional methods.
An adaptive motion selection algorithm for online hand-eye calibration was presented. It can adaptively set the thresholds of motion selection according to the characteristics of the unplanned motion sequence. It is a...
详细信息
An adaptive motion selection algorithm for online hand-eye calibration was presented. It can adaptively set the thresholds of motion selection according to the characteristics of the unplanned motion sequence. It is achieved by using polynomial-regression to predict the relationship between RMS error of calibration and thresholds. By using motion selection, the bad effect of small rotations and the degenerate motions such as pure translations will be removed and the accuracy of online hand-eye calibration be improved. Thus, this method can adapt itself to the online hand-eye calibration in various applications. Experiments using simulated and real data were conducted which present good results.
A system of map structure recognition and automatic map data acquisition was proposed. This system is based on binary skeleton image which is firstly obtained from scanned maps. Basic graph and super graph are propose...
详细信息
A system of map structure recognition and automatic map data acquisition was proposed. This system is based on binary skeleton image which is firstly obtained from scanned maps. Basic graph and super graph are proposed to facilitate the processing. Then kinds of interferential structures are analyzed and the corresponding removal methods are also introduced. Finally, the breaking point method is presented to polygonalize the house/building graph. The experiments conducted with various maps prove that the system is robust and effective to deal with complex scanned paper maps and can generate vector map automatically and correctly.
There exist uncertainties in rough set, and how to measure these uncertainties is a valuable problem to study. The basic methods for measuring uncertainty in rough set can not distinguish different granularity of part...
详细信息
There exist uncertainties in rough set, and how to measure these uncertainties is a valuable problem to study. The basic methods for measuring uncertainty in rough set can not distinguish different granularity of partitions. According to the definitions of accuracy of rough set and information entropy, this paper presented a new method for measuring uncertainty in rough set, and proved that this rough entropy monotonously increases while the partition granularity decreases. The new rough entropy can measure not only the size of uncertainty region in rough set but also the partition granularity. A practical examplt shows that this new method is effective.
A non-negative matrix factorization (NMF) based latent semantic indexing (LSI) model was introduced for image retrieval. Firstly, a semantic space is constructed using NMF-training algorithm. Then the hidden semantic ...
详细信息
A non-negative matrix factorization (NMF) based latent semantic indexing (LSI) model was introduced for image retrieval. Firstly, a semantic space is constructed using NMF-training algorithm. Then the hidden semantic features of the query image are extracted with NMF-testing algorithm. At last, ranking the query in this new semantic space and return some images to the user. The experiments show that the model provides better results than SVD-based LSI model and the one without LSI model.
This paper presented a new method to obtain the statistical relation between image features and matching probability. After integrating the Gabor wavelet features and some other parameters as matching area measures, t...
详细信息
This paper presented a new method to obtain the statistical relation between image features and matching probability. After integrating the Gabor wavelet features and some other parameters as matching area measures, this paper uses the support vector machine (SVM) classification method to transform the estimation of matching probability problem into a classifying one. The experiments show that the proposed method not only has faster computation speed than the method based on the correlation functions, but also gives a reasonable precise estimation.
暂无评论